Pairwise Interaction Probe¶
Interaction test — instantiates Composability Testing and Validation
Tests two components together — before either is trusted in any larger combination — to catch the interaction that neither reveals alone.
The smallest unit of composition is a pair, and the Pairwise Interaction Probe is the mechanism that exercises exactly that: it puts two components into the same environment, runs them together, and checks whether the thing each did correctly alone still holds when the other is present. Its defining move is isolation of the interaction itself — everything is held fixed except the fact that the two parties are now sharing a resource, an interface, a sequence, or a body. Where a unit test asks "does this component do what it promises?", the pairwise probe asks the strictly harder question "does the promise survive contact with that one neighbor?" It is the atomic experiment out of which every larger composability claim is eventually assembled, and it is deliberately narrow: it says nothing about triples, nothing about coverage strategy, nothing about release — only whether this two-way cell is clean.
Example¶
A hospital pharmacy is about to add a new oral anticoagulant to its formulary, and a patient on the ward is already taking a common macrolide antibiotic. Nobody doubts either drug on its own — each cleared its own trials. The open question is the pair. So the interaction service runs a pairwise check: it looks up each drug's contract (what enzyme pathway it is metabolized by, what it induces or inhibits, what it competes for), then asks whether co-administration violates any of them. The antibiotic, it turns out, inhibits the very metabolic pathway that clears the anticoagulant — so blood levels of the anticoagulant would climb, and the invariant "anticoagulation stays inside the therapeutic window" would break.
The probe's output is not "the drug is bad" and not "the antibiotic is bad." It is a single, localized verdict: this pair is unsafe as dosed, because A inhibits B's clearance. That verdict is precise enough to act on — substitute a non-interacting antibiotic, or reduce the anticoagulant dose under monitoring — without impugning either component elsewhere.
How it works¶
- Fix everything but the pairing. Run each component's own checks first so any failure in the combined run is attributable to the interaction, not to a pre-existing component defect.
- Read both contracts and look for the collision. Line up what one component emits, consumes, or changes against what the other assumes or depends on; the interaction lives wherever an output of one lands on an assumption of the other.
- Run the two together and judge against the invariant. Execute the combined pair in a controlled setting and check the post-conditions that must remain true — a boolean per-pair verdict: clean, degraded, or disqualifying.
- Localize the finding. Attribute the failure to the directional relationship (A affects B, or the reverse), because the fix usually attaches to one side of the pair.
Tuning parameters¶
- Directionality — treat the pair as symmetric (does A ⊕ B work?) or directional (does A harm B, and separately does B harm A?). Directional probes cost twice as much but localize the fix.
- Which pairs to run — all C(n,2) pairs, or only the pairs flagged as high-coupling. Exhaustive pairwise grows quadratically; skipping pairs trades cost for blind spots.
- Interaction strength threshold — how large an effect counts as a real interaction versus noise. Set it low and you drown in trivial couplings; set it high and you miss slow-building ones.
- Contact realism — how faithfully the shared environment reproduces the real one (same resource limits, same timing, same doses). Cheap synthetic contact misses interactions that only appear under real contention.
When it helps, and when it misleads¶
The pairwise probe's strength is precision and attributability: because only one relationship is under test, a failure points at a specific pair and usually a specific direction, which is exactly what the downstream fix needs. It is also the cheapest credible interaction evidence — quadratic in the component count, not exponential — so it is where composability testing almost always starts.
Its central failure mode is the seductive leap from pairwise-clean to safe in combination. Interactions are not additive: three components can conspire in ways no two of them do, so a suite of green pairwise checks can still sit atop a lethal three-way interaction[n1]. The classic misuse is polypharmacy review that clears each drug pair and declares the whole regimen safe, when the danger is a cumulative or three-way effect no pair exhibits. The guarding discipline is to treat pairwise-clean as a precondition, never a certificate: record that only two-way coverage was achieved, and hand the higher-order question to a coverage strategy that reasons about it explicitly.
How it implements the components¶
interaction_test_matrix— it populates and executes the two-way cells of the matrix, the pairwise layer every higher-order plan builds on.component_contract_inventory— it reads each component's declared assumptions and effects to predict where the collision will land, and localizes failures back to a contract line.invariant_and_safety_oracle— it judges each pair against the must-hold post-conditions, yielding a per-pair clean / degraded / disqualifying verdict.
It does not decide which combinations are worth running or bound the overall claim — the coverage budget and claim scope belong to Combinatorial Sampling Strategy — and it does not turn a failed pair into a durable rule; that triage belongs to Incompatibility Root-Cause Analysis.
Related¶
- Instantiates: Composability Testing and Validation — the pairwise probe supplies the atomic two-way evidence the archetype's coverage plan is assembled from.
- Sibling mechanisms: Combinatorial Sampling Strategy · Property-Based Composition Testing · Staged Integration Sandbox · Invariant Monitoring Dashboard · Fault-Injection Composition Probe · Incompatibility Root-Cause Analysis · Metamorphic Composition Test
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: Pairwise Interaction Probe operates as an active test, trial, simulation, drill, or rehearsal that generates evidence through a deliberate attempt or perturbation because it tests two components together — before either is trusted in any larger combination — to catch the interaction that neither reveals alone.
Independent corroboration: The frozen evidence defines Pairwise Interaction Probe as 'Tests two components together — before either is trusted in any larger combination — to catch the interaction that neither reveals alone', so its operative form is Experiment, Test & Rehearsal.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Testing two factors together for interaction is a canonical experimental-design method.
Related originating lineages:
- Computer Science & Software Engineering — Pairwise Interaction Probe also draws materially on computer science and software engineering's formal and practical treatment of computation, interfaces, data, and reliable systems, which shaped this mechanism rather than merely adopting it as an application.
- Engineering & Design — Pairwise Interaction Probe is most directly rooted in engineering and design's traditions of specification, testing, reliability, control, and physical-system construction. The lineage fits its defining practice: Tests two components together — before either is trusted in any larger combination — to catch the interaction that neither reveals alone.
Review resolution: Authoritative-source research resolves the primary-origin disagreement in favor of statistics experimental design. NIST/SEMATECH e-Handbook of Statistical Methods: Process Modeling documents the formative practice or theory represented here. The retained alternate domains identify material co-development or translation, while current applicability is recorded separately as domain_reach=multi_domain; origin_mode=convergent describes the historical relationship among lineages.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
[n1] A higher-order interaction is an effect present in a combination of three or more parts that is absent from every smaller subset — the statistical reason pairwise coverage cannot certify a whole. It is why "all pairs pass" and "the assembly is safe" are different claims. ↩